That's really a stricter requirement than idempotence (ie, functional
purity). I don't believe the compiler ever does this. The JVM might, if
it's able to inline some particular method call and discover therein that
no mutation happens, but that's probably pretty rare given the amount of
indire
You should then use memoize explicitly then. Of course, avoid this if
you have side effects in the function.
Luc
> In its optimization, does the Clojure compiler ever assume idempotency? That
> is, does it ever look at a function application `(f x y z)` and say to itself
> "I may freely substi
